Verifying the formatting of different renderings of a document for different locales presents significant challenges. For example, the number of different renderings may be too large for manual verification to be practical. Also, tools that detect differences between renderings at the pixel level are not practical, because pixel-level changes are expected for different renderings and thus do not provide a useful indication of potential formatting errors.